Social Representations of a dictatorial past and current support of authoritarianism

Abstract This study analyzed the structure of social representations of an authoritarian regime in Brazil (Military Regime, 1964-1985), the variables associated with reporting different emotions about that past, and support for authoritarian solutions in the present. A sample of 200 participants aged 18 to 85 years (M = 31.54; SD = 13.84) responded to a free word association task in which they indicated three words that came to mind about the Military Regime. A prototypical analysis showed a critical representation as the majority. However, positive representations and a mute zone effect (silence of controversial opinions) were found. Participants also answered several questions about the regime, the emotions they feel, and the support for its restoration. Regression analyses showed different variables predicting positive and negative emotions, and positive representations strongly predicting support for a current authoritarian solution.

An authoritarian military regime ruled Brazil for 21 years .Human rights violations has been attributed to state agents but the transition towards democracy was carried out within an Amnesty Law framework which has prevented any judicial prosecution.Only in 2012, a Truth Commission was carried out to investigate the human rights violations occurred during this period.Although it did not aim at conducting any prosecution, it pointed out the state agents and institutions responsible for 434 executions or disappearances and roughly 1800 cases of torture as well as other human rights violations (Comissão Nacional da Verdade, 2014).

Why some people follow authoritarian figures?
An authoritarian attitude could be encapsulated as an endorsement of a centralized power that does not accept the plurality and conflict of ideas that are characteristics of democracies (Cerutti, 2017).Authoritarian leaders promise to take decisive action on the problems in the name of the "people's will", opposing "the establishment", portrayed as corrupt or weak.A weaponized rhetoric (Pascale, 2019) against internal enemies that amplify resentments, suspicion of expert knowledge and the willingness for violence are other ingredients of authoritarianism (McCarthy, 2019).
Although studies have found associations of conservative political ideology with support of authoritarianism (see Duckitt, 2015 for a review on the theories of authoritarian personality), this relation was mediated by intolerance values and it is also related to low socioeconomic status (Napier & Jost, 2008).Periods of social and economic crisis are especially challenging to democracy foundations (Georgiadou et al., 2018).The perception of disintegration of the social system in a context of crisis is a powerful trigger to move people towards authoritarian beliefs, especially among poorer and less educated people who feels that they have lost control over their own lives (Chryssochoou, 2018).In addition to these explanations about the authoritarian attitudes, we postulate that the social representations of history play a crucial role on the support of current authoritarian positions in societies that had been ruled by dictatorships.

How people make sense of the past?
There are two main approaches to study how people collectively construct the knowledge about the past: the first is based on the concept of collective memory, and the second, on the social representations theory.
Both approaches present many points of contact being part of the social memory field (Mathias & Páez, 2018).
The collective memory paradigm has been established towards the idea that the past is a reconstruction of the events through the present frameworks (Halbwachs, 1950(Halbwachs, /2006;;Hirst et al., 2018).For instance, the generations who lived the military regime in Brazil would continuously reconstruct their memories to accommodate them in their current identities.However, younger generations cannot rely on any direct experience.They only have access to the memories that society keeps alive through the inter-generational transmission of stories, the registers of traces about the past or the historiography in textbooks.Therefore, younger generations do not remember but represent the past.That is the reason why the social representations theory is a helpful approach to study this phenomenon.

Social Representations Theory
The social representations theory analyze changes in the perception of social objects.It is especially useful to analyze how the expert knowledge is assimilated and transformed by common sense.The social representations are defined as: "a system of values, ideas and practices with a twofold function; first, to establish an order which will enable individuals to orient themselves in their material and social world and to master it; and secondly to enable communication to take place among the members of a community" (Moscovici, 2009, p.21).
Once created, the social representations serve as tools to understand, participate, and compete in the social arena (Moscovici, 2009).The transformation of a social representation seems to depend on the existence of a consistent minority that defies the established consensus, creating conflict and discussing the ideas once undoubted (Moscovici, 1979(Moscovici, /2011)).
The structural approach based on the central core theory (Abric, 2001) is a practical approach to investigate the social representations of polemical issues.It allows analyzing the relative contribution of the polemical elements to the overall social representation.Therefore, while the central core maintains the elements that are pivotal for the group identity, the periphery is more malleable.Social changes may trigger a transformation in the structure of social representation.The changes in elements on the central core usually indicate a transformation on the social representation meaning.However, studies on social representations of polemical issues might handle with social desirability, i.e., people sometimes do not speak out what they think.This phenomenon is known as the "mute zone" of the social representation and may require particular strategies of research (Abric, 2003).

Social representations of history
History, as any expert knowledge, is object of appropriation and transformation by the social representations processes.Social representations of history, therefore, can be considered as common sense theories about the past organized in the form of shared narratives providing a society a sense of continuity and indicating future paths related to its identity (Hilton &, 2017).
Although societies can reach broad consensus about what should be remembered or not (Liu & Páez, 2019), the creation of such agreement about the past can be tortuous because people differ in many social and individual aspects.The construction of different social representations of history may be related to current political issues (Snider, 2018), generation belonging (Muller et al., 2016;Sá et al., 2009), group identification (Obradovic, 2016), educational level (Sá et al., 2009) as well as political positioning (Muller et al., 2016).
Studies have shown that people had little information about the authoritarian regimes that occurred in the 20 th century in Brazil.For instance, about Getúlio Vargas' regime (1930)(1931)(1932)(1933)(1934)(1935)(1936)(1937)(1938)(1939)(1940)(1941)(1942)(1943)(1944)(1945), only 14% of adults and 58% of elders (who lived it) affirmed to have any information about it (Naiff et al., 2008).Education and displaying a political positioning, no matter which one, were associated with more information and to consider Getúlio Vargas as a dictator.Regarding the military regime , the social representations were found to be mainly negative, considering it a period of deaths, torture, censorship, and oppression (Sá et al., 2009).However, many young participants mentioned incorrect information such as the presence of a war.Also, there was a minority representation that considered the regime as a positive period of order and safeness.This positive representation was prevalent mainly in elders, conservative, and less educated people.Other studies with undergraduate students sample and using different techniques also showed an overall negative representation of the military regime (Weimer & Oliveira, 2020;Wolter et al., 2016).Social representations legitimizing the dictatorship as well as positive emotions on it were found in Argentina, but they were minority even among conservative participants (Arnoso et al., 2012).
We aimed, thus, at furthering the knowledge about the psychosocial variables associated with the social representations and emotions about an authoritarian past as well as its relations with the support for current authoritarian solutions.Firstly, we partially replicated Sá et al. (2009) study to check for changes in the structure of social representations of the military regime after the Truth Commission work.Then, we investigated several topics about the regime, especially the predictors of positive and negative emotions related to it.Finally, we also aimed at analyzing the relations of these variables and the social representations of the past as predictors of a current authoritarian position.
We posit the following hypothesis: H1: the central core of the social representations of the military regime will be more favorable to the regime than in Sá et al. (2009) study because of the recent rising of the conservatism in Brazil.That is, the frequency of positive words from the stimulus "military regime" will be higher, and these words will be mentioned firstly comparing to the results found by Sá et al. (2009).
H2: positive emotions about the military regime will be predicted by: older age, low education and conservative political opinions.The negative emotions about the military regime will present the opposite trend.
H3: agreeing with the military regime restoration will be predicted by the same variables predicting positive emotions plus a positive social representation of the regime.The positive social representation will present the higher effect.

Participants
The sample consisted of 200 Brazilian (54.3% women), from 18 to 85 years-old (M= 31.54,SD = 13.84).The study was approved by an Ethics Committee in Brazil (University of Bahia -CAAE: 54996416.4.0000.5686).Interviewers were volunteer university students trained in the application of the questionnaire.Data were collected between October 2016 and April 2017.Participants signed an informed consent letter that explained the study goals and guaranteed their anonymity.
Sampling was non-probabilistic, snowball type.Participants were contacted at their homes and places of work or study.There was a similar proportion of participants with low (high-school or less) (57.4%) and high (undergraduate and graduate) educational level (42.6%).Political positioning distribution was slightly center-left (M = 3.55, SD = 1.36, range 1-7) but this distribution is similar to general polls about this topic in Brazil (Corporación Latinobarómetro, 2020).Participants who affirmed being direct victims of the military regime were 5.8%, indirect victims (participants who had relatives or friends who were victims) 14.4% and no affected participants were 79.8%.

Materials
Preliminary questions: sociodemographic variables such as age, gender and level of education (elementary, high-school, undergraduate, and graduate).
Political Positioning: Likert-type scale from 1= extreme left to 7 = extreme right.
Political opinions: dichotomous questions about four social issues usually asked in surveys to define political positions in Brazil (Datafolha, 2017).The items asked participants whether they agree or disagree with: death penalty, reducing the legal age for criminal responsibility, legalization of marijuana and abortion.Disagreeing with marijuana and abortion legalization as well as agreeing with the death penalty and reducing the legal age for criminal responsibility scored for conservative.The opposite opinions scored for progressive.Then, we created a new variable for the sum of political opinions ranging from 1 = extreme progressive (four progressive opinions) to 5 = extreme conservative (four conservative opinions).
Information about the military regime: dichotomous item asking participants whether they "have ever heard about the military regime occurred in Brazil between 1964Brazil between -1985)).
Word association task: participants were asked to mention three words that come to mind when they think about the "military regime." Emotions associated with the military regime: Likert-type scale ranging from 1 = nothing at all to 7 = very much composed by two sub-scales, one for the positive emotions of hope, joy, and pride (α = .83)and the other for the negative emotions of anger, guilt, sadness, fear and shame (α = .83).
Opinions about the military regime and its restoration: two dichotomous items assessing whether or not participants agree that there were positive aspects of the military regime and whether or not they agree with a hypothetical military regime restoration.
Opinion about violations of human rights during the military regime: two dichotomous questions asking participants if there were violations of human rights during the regime and, in case of an affirmative answer, if the violations were legitimate.

Data Analysis
Data from the word association task were analyzed through a prototypical analysis conducted in the software IRAMUTEQ.This analysis allows investigating the central core of the social representation and their peripheral elements by visualizing them in four quadrants (Vergés, 1992;Wachelke & Wolter, 2011).The first (left-upper) representing the central core of the social representation, includes the elements most and firstly mentioned.The second and third quadrants (respectively right-upper and right-lower) represent the peripheral elements.They represent the elements mentioned lastly.Finally, the fourth quadrant (left-lower) depict the contrast zone elements.It comprises the evocations with low frequencies but evocated firstly.It means that these themes, although being a minority in the general sample, they are essential for those who evocated them.The corpus was built in the Open Office Calc software considering the order of evocation in order to analyze it through the matrix analysis method (Camargo & Justo, 2013).Similar words such as "United States" and "US" were patronized and words or expressions mentioned only once were excluded from the analysis.
We also carried out correlation, multiple and logistic regression analyses to investigate the variables related to positive or negative emotions on the authoritarian regime as well as to the opinion about its restoration.The words mentioned by each participant on the association task were categorized by three independent judges as positive, negative or neutral regarding the regime.There were no conflict on the categorizations.The neutral words were assigned as missing, and the variable named "positive social representations" was considered as dummy variable being 0 = negative and 1 = positive.

Information, evaluation, emotions and attitudes towards the military regime
A significant part of the sample (about a fourth) affirmed to have no information about the Brazilian military regime.About a third evaluated the regime somewhat positively, but only a minority agreed with its restoration (see Table 1).Likewise, opinions about the military regime were mostly critical.Most of the participants agreed with the idea that the regime violated human rights, and these violations were not legitimate.Accordingly, a paired sample t-test (t(129) = 8.09, p < .001)showed that negative emotions about the regime (M = 3.56, SD = 1.65) had a significant higher mean than positive emotions (M = 1.85,SD = 1.41).

Structure of the social representations of the military regime
Results from the prototypical analysis showed that the central core of the social representation of the military regime is mostly comprised of critical terms (see Table 2).However, the positive term "order" was found in the central core, indicating an outstanding minority representation.The first periphery is comprised only by negative terms regarding the military regime, but the second periphery showed the presence of positive words such as "respect," "peace" and "liberty" indicating that the positive representation of the military regime may be in construction from these elements plus the "order" found in the central core.Finally, the contrast zone presented a pattern similar to the central core.The majority of words were critical to the military regime, but positive words such as "good" were also found.

Evidences of a mute zone on the social representations of the military regime
The results found should be interpreted carefully.About a fifth of the sample was excluded from the analysis due to they were not able (or did not want to) mention more than one word regarding the military regime.Thus, we split the data from participants who mentioned two or three words (79.5%) from those who did not mention more than one word (20.5%) to check if there were systematic differences between them.Then, we conducted a correlation analysis to investigate the associations of missing the association task with psychosocial variables.Indeed, we found associations of not mentioning words regarding the military regime to most of the psychosocial variables investigated (see Table 3).Missing the association task was related to: young (r = -.17,p = .023),low educated (r = -.29,p < .001),conservative (r = .24,p = .002),positive emotions on the military regime (r = .21,p = .013)and agreement with its restoration (r = .32,p < .001).Thus, we should be careful in analyzing the results found for the structure of the social representation, remembering that they display only those who were willing to talk about this subject.In the same way, it may exist an active minority of supporters trying to change the overall critical social representations of the military regime, it also seems to exist passive people who tend to silence themselves about this topic but to assimilate the positive representations.

Psychosocial variables related to a positive view of Brazilian military regime
We conducted correlation analyses (Pearson and point bi-serial) to investigate the psychosocial variables associated with different opinions, emotions and social representations regarding the military regime.First, we checked whether political self-identification was similar to political opinions.Associations were not as high as we could expect (see Table 3).Thus, we checked the frequencies on crosstabs.The results showed that few participants who had progressive opinions about social issues identified themselves as right-wing (5%).However, many participants who identified themselves as left-wing had conservative opinions (32%).Some participants may have identified themselves thinking more about economic than social issues.Consequently, we chose to report the analyses considering the political opinions rather than the political self-identification due to our questions are not related to economy.Although the direction of the results did not change (see Table 3), the effects were stronger using the last variable, and also we were able to detect some associations we could not do using the political self-identification.
Results from the correlation analyses showed weak associations of gender and age but greater associations of education and political opinions with the variables regarding the military regime (see Table 3).Being older showed a positive association with reporting positive emotions on the military regime.On the other hand, higher educational level and progressive opinions were positively associated with negative emotions on the military regime, whereas, lower education and conservative opinions were related to positive emotions and social representations about the military regime, as well as to agreeing with its restoration.
It is important to highlight that higher education was positively associated with progressive opinions, but surprisingly, this result was not found for political self-identification.Moreover, although the results for emotions were similar (political opinions presenting stronger effects than self-identification), right-wing self-identification was not associated to agreeing with the restoration of the military regime, while conservative opinions did.These results call into question whether the wide use of the political positioning scale in a single Likert-type item is an appropriate measure of participants political view, especially while studying polemical issues.
In order to better understand the results, we conducted a linear multiple regression analysis to investigate which psychosocial variables predict the positive and negative emotions regarding the military regime.We tested a model on the Enter mode including the three variables found in a previous study (Sá et al., 2009) as associated to the social representations of the military regime: age, political opinions (replacing selfidentification), and education.Assumptions were met, and significant regression equations were found for positive and negative emotions (see Table 4).Different variables predicted positive and negative emotions on the military regime.While low education showed significant effects on the positive emotions (older age showed significant effects at p < .10),only progressive opinions was a significant predictor of negative emotions on the military regime.

Are the social representations of history the key variable in supporting a current authoritarian solution?
Finally, we conducted a logistic regression analysis to investigate the effects of the psychosocial variables on agreeing with the military regime restoration.We tested a model including as predictors age, education and political opinions, but also positive social representations of the past regime to investigate the specific effects the social representations of the past have on the current authoritarian attitude.Assumptions were met, and a significant regression equation was found (see Table 5).Positive social representations and low education were found as significant predictors of Note.* p < .05,** p < .01,*** p < .001.Gender was coded as 1 = male, 2 = female.Education was coded as 1 = elementary, 2 = high-school, 3 = undergraduate, and 4 = graduate.Political opinions ranged from 1 = extreme-progressive to 5 = extreme-conservative.Missed association task was coded as 0 = did not answered or mentioned only one of three words and 1 = mentioned two or three words.Social Representation Dictatorship was coded as 0 = negative, and 1 = positive.Restoration Regime was coded as 0 = disagree, 1 = agree.agreeing with the regime restoration (conservatism was also found as a significant predictor at p < .10).This result suggests an important relationship between the social representations of history and current support of authoritarian solutions.Still considering the effects of other variables such as age, conservatism and low education, those who has a positive social representation of the military regime in Brazil are 15 times more likely to agreeing with an authoritarian solution than those who holds a negative social representation.

Discussion
This study had three main goals: first, to examine transformations on the social representations of history about an authoritarian past in Brazil after a Truth Commission work; second, to investigate the psychosocial variables related to positive and negative emotions regarding the regime, and third, to analyze whether these psychosocial variables and positive social representations about the regime would predict agreeing with its restoration.
Hypothesis 1 positing that the central core of the social representations of the military regime would be more favorable to it was not confirmed.Results from the prototypical analysis presented an overall similarity with those found by Sá et al. (2009).Although a revisionist perspective which represents the authoritarian past as a positive period of order, discipline, respect, and even liberty has been increasing, the social representations of the military regime seem to be still mainly critical.However, the term "order" displayed in the Brazilian national flag was part of the central core, indicating that a nationalistic representation about the past may be rising.
Yet, the results suggest that some people who hold positive representations about the regime silence themselves, not mentioning words on the association task.Specifically, we found that low education, positive emotions on the regime, and agreeing with the military regime restoration were associated with not mentioning words.This result may point out the existence of a mute zone on the social representation of the military regime in Brazil, that means, a representation that is not expressed because it is counter-normative, although they do exist and influence people's behavior (Abric, 2003;Menin, 2006).These studies about the mute zone are usually about prejudice.Finding it in a study about social representations of history can open up new paths on this field.Further studies may take into account special techniques and cautions to assess these kinds of contents reducing normative pressure.
Confirming H2, positive emotions on the military regime were positively associated with age and conservative opinions, and negatively associated with education.However, when we include these variables  = 5.24 (1.65-16.64),Wald t = 7.90 p = .005)meaning that the probability of people who did not answered the association task to endorse an authoritarian solution is 5 times higher than for those who answered.That is an evidence for a mute zone effect on the social representations of an authoritarian regime.
together as predictors in a multiple regression model, only lower education and older age (at p < .10)remain as significant predictors of positive emotions on the military regime.These results were somewhat similar to those found by Sá et al. (2009) indicating that the emotions about the past could be a dimension aligned with the social representations of history.Indeed, our results showed high correlations between positive social representations and reporting positive emotions on the regime.Although political positioning is considered as an essential variable on representing the military regimes in Latin America (Muller et al., 2016), a special care should be taken about the political positioning variable.Our study showed limited associations between selfpositioning and political opinions.
Our results indicate that education is also an important variable in analyzing the social representations of history.Educational level has stronger effects on political knowledge in unequal societies such as Brazil (Grönlund & Milner, 2006).Low education was associated with considering the former authoritarian president Getúlio Vargas (1930Vargas ( -1945) ) as not being a dictator (Naiff et al., 2008) and were also related to positive representations of the military regime (Sá et al., 2009).Although a significant part of our sample affirmed to have some information about the military regime, still a fourth have not ever heard about such an important period of Brazilian history, showing a lack of education and public discussing about this subject.Accordingly, a study on the collective memory of lived events showed that the military regime was not mentioned as an important social event even by the people who lived it, differently from Argentina, Chile or Uruguay where the dictatorship are among the most mentioned ones (Páez et al., 2017).
The associations of age to positive emotions on the past regime provide support for the idea that the generational cohorts hold different collective memories, emotions or representations of the past because some people have direct experience to rely on whilst others have indirect knowledge through formal education or oral transmission (Halbwachs, 1950(Halbwachs, /2006)).Moreover, the effects of age on positive emotions regarding an authoritarian past are in line with previous findings (Muller et al., 2016) indicating that those who lived the events hold qualitatively different representations than those who did not.
Regarding the negative emotions on the regime, we did not find fully support for H2.Contrary to H2, age and education did not show significant associations with negative emotions on the military regime.Confirming H2, however, progressive political opinions were associated with reporting negative emotions on the military regime.Regarding age, an explanation could be that the negative representations of the military regime are more consensual than the positive, being shared more equally across generations, while the positive representations are restricted to nostalgic conservative elders.Regarding education, it seems to prevent the developing of positive emotions, but not necessarily increasing the negative ones.These results seem to be in line with a previous study showing that as the time passes, descriptive non-emotional representations are expected (Naiff et al., 2008).
The results provided support for the idea that holding positive social representations about the authoritarian past could represent a threat to democracy.According to H3, mentioning positive words on the association task was the main variable predicting the agreement with the authoritarian regime restoration (low education and conservatism showing more limited effects).However, contrary to H3, we did not find significant effects of age.
It is important to note that agreeing with the military regime restoration is a minority position (12.5% of the sample).This proportion is similar to that found in polls about the preference for authoritarian regimes (Corporación Latinobarómetro, 2020).Therefore, we could expect this position to be held by radicalized individuals.Our results suggest that, although most of those who agree with the regime restoration are conservative, this position is restricted mainly to those conservative who hold positive social representations of the authoritarian past, which is minority, even among conservatives.

Concluding Remarks
This study investigated the structure of social representations of the Brazilian military regime , the variables associated with reporting different emotions about that past, and the support for a hypothetical return of a military rule.
Even though we did not detect, as our hypotheses suggested, a significant increase in the positive representations about the regime when we replicated the Sá et al. (2009) study, we do found evidence of a mute zone effect (silence of polemical opinions).This result calls into question the existence of a portion of the population with authoritarian attitudes that are not detected through traditional survey techniques.
The recent rise of populist authoritarian positions may be explained by both, the context of social crisis (Georgiadou et al., 2018) or a cultural backlash (Inglehart & Norris, 2016).Our results seem to limit the interpretation of a cultural backlash that associates the populist/authoritarian rising to nostalgic elders, once age was not associated to neither a positive social representation of the military regime nor agreeing with its restoration.Probably, societies with an authoritarian past have a great booster of current authoritarian attitudes on the idealized social representations about this period even among the younger people.Therefore, alongside with political positioning and opinions, education and diffusion of information about the past seems to be important variables to take into account.Indeed, we found that low education is the main variable related to reporting positive emotions about the authoritarian past, while political opinions are more related to report negative emotions.Thus, our results suggest that positive and negative social representations about an authoritarian past may operate through different mechanisms, respectively, education vs political identification.
Finally, another goal of this study was to investigate the effects of positive social representations about the regime on the support for a new democratic rupture.It may seem obvious that people who wish a return of the military regime are those who represent the military regime positively.However, we aimed at examining the effects of these representations in relation to other variables, i.e. conservatism, older age, and lower education in predicting current authoritarian positions.Our results suggest that, even if low education (or low socioeconomic status) are important predictors of support for authoritarianism (Chryssochoou, 2018;Napier & Jost, 2008), in contexts with a dictatorial past, it is the active construction of positive representations about this past that relates more strongly to support for a new democratic rupture.In other words, it does not seem to be lack of information, but misinformation about the past (the construction of a representation of an idyllic paradise of order, security, and even freedom), that fuels a wish for the restoration of that alleged glorious past.
This study presented only a first glance at the different variables which could be related to the social representations of an authoritarian past.It was limited in some aspects.Particularly, the sample was quite small and contained many missing responses.Therefore, we might have failed to detect some small effects.Future studies may investigate the social representations of history in relation to other political objects such as the ideas of democracy and authoritarianism.They could also investigate whether different levels of political knowledge and engagement are associated with supporting democracy or authoritarianism.

Table 2 .
Quadrants For The Prototypical Analysis Regarding The Stimulus "Military Regime"

Table 3 .
Correlation Matrix For The Variables Related To The Social Representations Of The Military Regime

Table 5 .
Logistic Regression Of Agreement WithThe Restoration Of The Military Regime p < .05,** p < .01.Education was included as a dichotomous factor (0 = university or technical; 1 = high-school or less).92% of the data was correctly predicted by the model.Sensitivity (authoritarian endorsers correctly predicted by the model): 46%; Specificity (nonauthoritarian correctly predicted by the model): 97%.When miss word is included in the model instead of positive representations, it is still significant (χ 2 : 30.15, p < .001;Nagelkerke R 2 = .35;Odds Ratio Note. † p < .10,*